Design, synthesis and application of carbazole macrocycles in anion sensors

Article abstract of DOI:10.3762/BJOC.16.157

Carboxylate sensing solid-contact ion-selective electrodes (ISEs) were created to provide a proof-of-concept ISE development process covering all aspects from in silico ionophore design to functional sensor characterization. The biscarbazolylurea moiety was used to synthesize methylene-bridged macrocycles of different ring size aiming to fine tune selectivity towards different carboxylates. Cyclization was achieved with two separate strategies, using either amide synthesis to access up to –[CH₂]₁₀–macrocycles or acyl halides to access up to –[CH₂]₁₄–macrocycles. Seventy-five receptor–anion complexes were modelled and studied with COSMO-RS, in addition to all free host molecules. In order to predict initial selectivity towards carboxylates, ¹H NMR relative titrations were used to quantify binding in DMSO-d₆/H₂O solvent systems of two proportions – 99.5%:0.5% m/m and 90.0%:10.0% m/m, suggesting initial selectivity towards acetate. Three ionophores were selected for successful sensor prototype development and characterization. The constructed ion-selective electrodes showed higher selectivity towards benzoate than acetate, i.e., the selectivity patterns of the final sensors deviated from that predicted by the classic titration experiments. While the binding constants obtained by NMR titration in DMSO-d₆/H₂O solvent systems provided important guidance for sensor development, the results obtained in this work emphasize the importance of evaluating the binding behavior of receptors in real sensor membranes.

Full text of DOI:10.3762/BJOC.16.157

Design, synthesis and application of carbazole macrocycles
in anion sensors
Alo Rüütel1, Ville Yrjänä2, Sandip A. Kadam1, Indrek Saar1, Mihkel Ilisson1,
Astrid Darnell1, Kristjan Haav1, Tõiv Haljasorg1, Lauri Toom1, Johan Bobacka2
and Ivo Leito*1
Full Research Paper
Open Access
Address:
Beilstein J. Org. Chem. 2020, 16, 1901–1914.
1Institute of Chemistry, University of Tartu, Ravila 14a, 50411 Tartu,
doi:10.3762/bjoc.16.157
Estonia, https://analytical.chem.ut.ee and 2Johan Gadolin Process
Chemistry Centre, Laboratory of Molecular Science and Engineering,
Åbo Akademi University, Biskopsgatan 8, FI-20500 Turku/Åbo,
Finland
Received: 21 April 2020
Accepted: 16 July 2020
Published: 04 August 2020
Email:
This article is part of the thematic issue "Molecular recognition".
Guest Editor: J. Niemeyer
Ivo Leito* - ivo.leito@ut.ee
*
Corresponding author
©
2020 Rüütel et al.; licensee Beilstein-Institut.
Keywords:
License and terms: see end of document.
anion sensors; carboxylates; ionophores; macrocycles; sensor
prototype
Abstract
Carboxylate sensing solid-contact ion-selective electrodes (ISEs) were created to provide a proof-of-concept ISE development
process covering all aspects from in silico ionophore design to functional sensor characterization. The biscarbazolylurea moiety was
used to synthesize methylene-bridged macrocycles of different ring size aiming to fine tune selectivity towards different carboxyl-
ates. Cyclization was achieved with two separate strategies, using either amide synthesis to access up to –[CH2]10– macrocycles or
acyl halides to access up to –[CH2]14– macrocycles. Seventy-five receptor–anion complexes were modelled and studied with
COSMO-RS, in addition to all free host molecules. In order to predict initial selectivity towards carboxylates, 1H NMR relative
titrations were used to quantify binding in DMSO-d6/H2O solvent systems of two proportions – 99.5%:0.5% m/m and
9
0.0%:10.0% m/m, suggesting initial selectivity towards acetate. Three ionophores were selected for successful sensor prototype
development and characterization. The constructed ion-selective electrodes showed higher selectivity towards benzoate than
acetate, i.e., the selectivity patterns of the final sensors deviated from that predicted by the classic titration experiments. While the
binding constants obtained by NMR titration in DMSO-d6/H2O solvent systems provided important guidance for sensor develop-
ment, the results obtained in this work emphasize the importance of evaluating the binding behavior of receptors in real sensor
membranes.
Introduction
In 2013, Otto S. Wolfbeis asked the supramolecular commu- new actual chemical sensor systems has been slow. We have
nity a question of justified critique: why do we have so few true seen numerous potentially very promising receptor candidates
sensors? [1]. Indeed, in the recent decades the progress towards from synthetic chemistry – essential as the central recognition
1901

Beilstein J. Org. Chem. 2020, 16, 1901–1914.
elements of chemical sensors. However, the research typically (HBD) groups can be added with substituents. The solubility of
ended with demonstration of the ability of analyte binding. such receptors can be tuned using functional groups in the 3,6-
Rarely have the receptors found their way into functioning positions of the carbazole. High lipophilicity is achievable by
chemical sensor prototypes. The reason for this is that the step using alkylation and leads to the possibility of using the recep-
from a well-binding receptor to a sensor prototype poses signifi- tor dissolved in the lipophilic polymeric sensor membrane,
cant challenges.
without the need for chemical linking. Cyclization of the recep-
tor is possible via the positions 8 of the carbazole, using, e.g.,
The development of a chemical sensor consists of a number of amide bonds.
steps that require knowledge and skills from different disci-
plines. Although not always necessary [2], specific selectivity
towards a particular analyte is accessible with the use of a
specialized ionophore.
We directed our efforts towards sensing carboxylates. The
detection of these analytes is important in a number of areas and
there is an obvious lack of easy-to-use sensing methods.
Carboxylates are challenging analytes for two reasons. Firstly,
binding is mainly achieved through either ion–ion interactions
Figure 1: The biscarbazolylurea moiety.
or hydrogen bonding. The primary site of interaction is the
X-COO− group. The geometry of the carboxylate group (and to
a large extent also its charge distribution) is the same for all Anion receptors containing carbazole and amide functionalities
carboxylate anions. Therefore, high affinity towards carboxyl- were investigated in numerous works [6-9]. In some cases,
ates is easily achievable, but selectivity towards a specific these functionalities were incorporated into macrocyclic
carboxylate anion will prove challenging.
systems, thereby offering valuable insight for design criteria.
For example, a carbazole-urea macrocycle was reported previ-
Secondly, the X-COO− group itself may become protonated ously [10], however, the binding of anions occurred outside the
under specific conditions. In this scenario, selectivity would be receptor due to modest dimensions of the macrocyclic cavity.
lost immediately. This puts restraints on the molecular design of Using click-chemistry, a carbazole-triazole macrocycle,
the receptor molecule. The host molecule cannot include any “tricarb”, was prepared that showed the ability to form non-
functional groups that would allow the analyte to become proto- covalent superstructures [11].
nated.
The rings should be of remarkable size for accommodating even
In order to incorporate both functional parts of a carboxylate, a small anions, such as acetate. In many cases the receptor is for-
macrocyclic receptor architecture is desirable. By using a cyclic mally a macrocycle, but the anion is bound on its “surface”, not
structure, it is possible to accommodate a solvophobic environ- inside the ring [10,12]. A well-known example is the
ment alongside polar functional groups. The benefits of this calix[4]pyrrole family, which emerged as an attractive neutral
protein-inspired architecture have recently been praised in the host for anions [13]. It is formally a 16-membered ring, but for
use of naphthotubes [3]. Association between the receptor and the above reasons no anion, not even F− fitted inside the ring –
the analyte may change considerably when the receptor is at- all bound anions reside on the surface of the ring [14]. Often in
tached to an electrode surface or embedded in a sensor mem- such cases, two receptors stack the guest molecule in a “sand-
brane. Water, as the most common environment for real-life wich” like manner, e.g., the “cyanostar” macrocycle [15,16].
sensing, also influences binding [4]. Within an ion-selective Although not a macrocycle, a successful binding motif that is
electrode (ISE), lipophilicity is mostly contributed by the low noteworthy for its ability to bind carboxylates in polar environ-
polarity membrane, which is commonly based on plasticized ments, is the guanidiniocarbonylpyrrole (GCP) moiety de-
PVC [5].
signed by Carsten Schmuck and Michael Schwegmann [17].
One possible receptor family for carboxylates is based on the The transition from complexation studies to the next phase of
biscarbazolylurea moiety (Figure 1). This moiety fulfils several constructing a prototype sensor is where progress often stopped
key design requirements. At least four hydrogen bonding sites and one answer to Wolfbeis’ question can be found. If we ever
(
N–H groups) are available and positioned in a favourable ge- wish to see the developed champion molecule in commercial
ometry for carboxylates. Additional hydrogen-bond-donor use, several conditions must be met. The yield of the synthetic
1902

Beilstein J. Org. Chem. 2020, 16, 1901–1914.
pathway can be low and require optimization. The chemical self-plasticizing. Plasticized poly(vinyl chloride) (PVC), which
characteristics of the champion molecule might exclude its suc- was used in this study, is an example of a commonly used mem-
cessful commercialization – solubility issues, chemical insta- brane matrix.
bility, secondary effects during binding (e.g. superstructures),
too high affinity constants, insufficient selectivity towards the Permselectivity is conferred to the polymeric membrane by the
analyte, etc. If a champion with sufficient characteristics is addition of an ion exchanger, which is a lipophilic ion, and an
found, then it can be incorporated into a prototype sensor.
ionophore. The ion exchanger provides permselectivity for
anions or cations depending on whether the ion exchanger salt
Different strategies are available for electrochemical sensor de- contains a lipophilic cation or anion, respectively. Additional
velopment [18,19]. While several cation-sensing electrodes selectivity is provided by the ionophore, which is typically
have been successfully commercialized, only a limited number confined to the membrane in one of two ways. One possibility
of anion-sensing electrodes are commercially available. This is is to immobilize a receptor in a polymeric membrane with the
largely due to the challenges associated with the selectivity of use of covalent bonding. Another strategy is to dissolve the re-
ionophores towards anions [20,21]. In this work, potentio- ceptor molecule in the polymeric membrane, which was the ap-
metric solid-contact ion-selective electrodes (SC-ISEs) were proach that was used in this study.
chosen as the sensor type for several reasons. Potentiometric
ISEs possess several sought-after characteristics for a sensor: In this case, the receptor molecule must be highly lipophilic. It
easy to produce, low manufacturing and operating costs, easy to is usually easier to tune the solubility of the receptor molecule
use, portable, and sufficiently quick response [5,22,23]. than it is to add the functional groups necessary for covalent im-
SC-ISEs have been used for decades in research and can be mobilization [25]. Therefore, these types of sensors are easier to
used to make less fragile and more easily miniaturized ISEs prepare but their life span is shorter due to the possibility of the
than the conventional ISEs [24]. Other electrochemical readout receptor molecule leaking from the polymeric membrane into
principles can be used with these sensors, but potentiometry has the analyte solution. The ratio of ionophore to ion exchanger is
well established theoretical foundations and protocols for the an important parameter and typically an excess of ionophore
characterization of ISEs. Equilibration times are typically also (e.g., 2:1) is used to prevent coextraction of ions with the oppo-
short in potentiometry. This is appropriate for the current study site sign from the sample along with the primary ion. The con-
where the life span of the sensors may turn out to be shorter struction of the SC-ISEs used in this work is shown schemati-
than expected.
cally in Figure 2. In these electrodes, poly(3,4-ethylene-
dioxythiophene) doped with chloride (PEDOT-Cl) was used as
The potentiometric response of an ISE to the activity of the pri- the solid contact (ion-to-electron transducer).
mary ion, i, can be described by the Nikolsky–Eisenman equa-
tion (Equation 1), which expands upon the Nernst equation by
accounting for the interference caused by the activities of inter-
fering ions, j, with a potentiometric selectivity coefficient,
.
(1)
The selectivity of SC-ISEs is typically achieved by coating the
ion-to-electron transducer with an ion-selective membrane
(
ISM) containing the ion receptor (ionophore). The ISM matrix
can be made of several different types of materials such as
glass, crystal, or polymer. pH electrodes with glass membranes
are the most used ISEs and produce highly selective responses
over a wide range of concentrations. Polymeric membranes
have gained popularity due to their robustness, the ability to
fine-tune their properties and the ability to target a larger variety
of ions compared to e.g. glass. Plasticizers may be required to
improve transmembrane diffusion rates unless the polymer is
Figure 2: The structure of the solid-contact ion-selective electrode
sensor): a) glassy carbon as the electronic conductor; b) insulating
shell made from PVC; c) PEDOT-Cl as the solid contact (ion-to-elec-
tron transducer); d) an ion-selective membrane with a plasticized PVC
matrix. The inset shows a cross-section of the center of the sensor tip.
(
1903

Beilstein J. Org. Chem. 2020, 16, 1901–1914.
Results and Discussion
Design
expected that these macrocycles will have lower affinity con-
stants.
We aimed to investigate the effects of ring size on the binding
of carboxylates and the biscarbazolylurea moiety was selected. MC002 was of sufficient size to accommodate smaller carbox-
The studied receptors are shown in Figure 3. In order to achieve ylate ions. However, in the most stable computed structures the
macrocyclization, we decided to use linear aliphatic dicar- hydrogen-bond donor NH fragments directed towards the
boxylic acid residues (from glutaric to thapsic acid) with differ- outside of the ring. In addition, the increased ring size could
ent chain lengths, allowing manipulation of the ring size. Simi- accommodate solvent molecules. Thus, an additional hindrance
lar strategies for closing macrocycles have previously been to anion binding could occur due to the need to displace solvent
published [26-28].
molecules.
To predict possible binding properties, all methylene-bridged According to the computations, the rings formed by macro-
macrocycles and their binding to several carboxylates were cycles starting from MC006 had sufficient size to fit anions,
modelled computationally using COSMO-RS. Several observa- starting with formate. MC008 was able to fit acetate and lactate.
tions were made from the computational structures.
Thus, an additional binding site was available for lactate. In the
complexes of lactate with MC006–MC009, a hydrogen bond
Intramolecular hydrogen bonds between the urea carbonyl and was present between the oxygen of the hydroxy group of lactate
the carbazole NH protons were present in macrocycles MC001 and the NH group of the macrocyclic amide. See Figure 5 for
and MC003, as indicated in Figure 4. The rings of these macro- visualization. An equilibrium likely existed between two
cycles, alongside with MC004 and MC005, were too small to conformers where one complex had an additional hydrogen
bind any of the studied anions. In the case of MC001, two bond and the other included intramolecularly bonded
anions could interact with it in such a way that binding occurs lactate. For the lactate complexes with the larger receptors,
outside the macrocycle. For MC003–MC005, the ring could whether one dominant conformer or a mixture of two
accommodate the carboxylate group COO−, however, the conformers was preferred seemed to be dependent on the
residue –X would be forced out of the macrocycle plane. It was –CH2– linker length.
Figure 3: Studied receptor molecules.
1904

Beilstein J. Org. Chem. 2020, 16, 1901–1914.
Figure 4: MC001 and MC003 lowest energy conformers (COSMO-RS) showing intramolecular bonds. Color coding: white – hydrogen; green – car-
bon; blue – nitrogen; red – oxygen.
Figure 5: a) Complex of MC008 with acetate; b) complex of MC006 with formate; c) complex of MC007 with lactate showing an intramolecular hydro-
gen bond; d) complex of MC009 with lactate showing an additional hydrogen bond with an amide NH of the macrocycle.
Benzoate and pivalate were too large to fit the rings of even the Synthesis
biggest macrocycles. However, the out of plane residue –X The general synthesis route is presented in Scheme 1. It was to
should be able to, at least to some extent, benefit from the some extent similar to the one that has previously been
solvophobic effects of the –CH2– linker if binding occurred in published by Sanchez et al. and Liu et al. [6,29] (see below).
water (and possibly also in DMSO).
The synthetic route relied on the initial alkylation of the 3 and 6
1905

Beilstein J. Org. Chem. 2020, 16, 1901–1914.
Scheme 1: The synthetic pathway to receptors CZ016 and MC001–MC014. The reaction yield for 2–3a/3b is given as a sum for two compounds, as
they were inseparable by purification.
1906

Beilstein J. Org. Chem. 2020, 16, 1901–1914.
positions of carbazole for tuning the solubility profile and for Binding measurements
directing reactivity towards the 1 and 8 positions. The obtained The binding properties of the macrocycles are expressed by the
molecule 2 was then nitrated and hydrogenated to obtain the logarithms of the binding (association) constants logKass, which
amine 5a. The amine group was protected with a Boc-group at were measured using our previously published relative NMR
one substituent and coupled using CDI, affording the protected method under fast exchange conditions [30]. Analogous to our
biscarbazolyl derivative 7. TFA was used for deprotecting 7, earlier works with similar receptors, a binding stoichiometry of
which led to the corresponding free amine derivative 8 that 1:1 was ensured by using low concentrations of receptor and
could directly be used for cyclization.
anion – mimicking the real situation when using the receptors in
sensors. The stoichiometry was confirmed by data treatment –
We were unable to successfully follow all steps in the synthesis the experimental results agreed with the 1:1 binding model.
scheme by Sanchez et al. [6]. The problematic step was the Please refer to Supporting Information File 1 for additional
nitration reaction. Due to the highly acidic environment partial details on the relative binding measurements. The binding
loss of one of the tert-butyl groups in 2 took place. This consid- affinities for MC002 with lactate and pivalate were obtained
erably reduced the yield and resulted in a mixture of products by UV–vis measurements as absolute values [31]. Five carbox-
3
a and 3b that was challenging to separate due to similar prop- ylates were studied in the form of their tetrabutylammonium
erties. Careful optimization of the nitration process did not lead salts: pivalate, acetate, benzoate, lactate, and formate. The
to acceptable results. Therefore, we decided to modify the reac- anions were titrated in DMSO-d6/H2O solvent systems of two
tion scheme. After alkylation, compound 2 was instead bromi- proportions – 99.5%:0.5% m/m [30] and 90.0%:10.0% m/m
nated, followed by amination of 4 leading to the desired [31].
diamino product 5a. Amine protection and subsequent coupling
procedures, leading to 8, were the same as described in refer- The 0.5% m/m water addition was necessary to make the results
ence [6].
more stable by minimizing the effect of DMSO’s hygro-
scopicity and thereby stabilizing the water content in the solu-
It is worth noting that the described cleavage problem during tion. Since anions are solvated considerably more strongly
nitration was specific to tert-butyl groups. We also studied in pure water than in DMSO, we also studied binding
the possibility of hexyl substituents instead of tert-butyl using a higher water content in DMSO. The 10.0% m/m
groups [25]. In this case, the nitration process yielded only mixture was expected to lower the binding affinities for all re-
the desired product. However, since this possibility was ceptor–anion combinations. The results are presented in Table 1
investigated after the production of the first macrocycles bear- and Figure 6.
ing tert-butyl substituents, we decided to not switch substitu-
ents.
Initially, the trend suggested a steady increase of the binding
affinity with increasing ring size. This, however, reached a
The formation of macrocycles from 8 was achieved following plateau at around 7–8 carbon atoms in the chain and the
two different synthetic strategies. For receptors maximum affinity value for any given anion did not increase
MC003–MC010, amide bond formation was carried out using thereafter, except for the case of pivalate (MC009 had the
dicarboxylic acids alongside EDC·HCl, HOBT and DMAP/ highest logKass). With chain lengths of more than
DIPEA. This method was accompanied with a systematic de- 9 carbon atoms the logKass values started to decrease. The
crease of the yield with the lengthening of the –CH2– methy- initial increase of the binding affinity suggested that just
lene bridge. To counteract the decrease of yield a second ap- fitting into the ring was not a sufficient criterion for predicting
proach was adapted for the larger receptors MC011–MC014. binding affinity. This was because for the macrocycles
For this the corresponding dicarboxylic acids were first con- MC003–MC005, the computational geometries suggested that
verted to the diacyl chlorides, which were then used for cycliza- the ring is too small to bind any of the studied anions. Addition-
tion. This mitigated the yield drop and provided access to the ally, for MC003, intramolecular hydrogen bonds could be
higher macrocycles MC011–MC014.
present.
In addition, three other receptor molecules were prepared. For most anions, the champion molecules were MC007 and
MC001 and MC002 were prepared using CDI coupling of the MC009. According to the computation results, neither benzoate
corresponding amino derivatives according to Scheme 1. The nor pivalate were able to fit into the ring of these receptors
synthesis of CZ016 is also described in Scheme 1. Details of all because these anions were too large. Yet, MC007 and MC009
synthetic procedures can be found in Supporting Information bound benzoate and pivalate better than larger macrocycles
File 1.
where the steric strain was expected to be lower.
1907

Beilstein J. Org. Chem. 2020, 16, 1901–1914.
Table 1: Binding affinities logKass of different carboxylates in 0.5% and 10% DMSO-d6/H2O solvent systems.a
anion
receptor
pivalate
.5%
acetate
0.5%
benzoate
0.5%
lactate
0.5%
formate
0.5%
0
10%
10%
10%
10%
10%
CZ016
MC001
MC002
MC003
MC004
MC005
MC006
MC007
MC008
MC009
MC010
MC011
MC012
MC013
MC014
5.16
3.53
3.46
3.69
4.40
4.93
4.90
5.64
5.23
5.82
5.21
5.43
5.40
5.44
5.48
b
b
b
b
4.82
3.33
3.30
3.50
4.47
5.00
5.17
5.69
5.34
5.69
5.36
5.18
5.00
4.97
4.97
b
b
b
b
4.20
2.84
2.86
2.93
3.69
4.17
4.47
4.91
4.64
4.95
4.58
4.62
4.41
4.33
4.32
b
b
b
b
3.29
2.26
2.68
2.40
2.99
3.36
3.47
4.04
3.62
4.07
3.69
3.78
3.62
3.58
3.55
b
b
b
b
3.81
2.82
2.66
2.68
3.48
4.06
4.51
4.80
4.63
4.59
4.18
4.00
3.86
3.76
3.77
b
b
b
b
2.06
3.63
b
1.97
3.78
b
1.63
3.80
b
1.75
4.03
b
1.69
3.58
b
4.12
3.86
3.70
3.07
3.26
3.80
3.75
3.67
3.55
3.55
4.60
3.74
3.73
2.73
2.84
b
b
b
b
b
b
b
b
b
b
b
b
b
b
b
b
b
b
b
b
b
b
b
b
b
aAll anions were used as tetrabutylammonium salts at ambient temperatures. The standard uncertainties of the logKass values, accounting for consis-
tency of measurements, uncertainty of anchoring and possible systematic effects (see reference [8] for details) were in the range of 0.04 and 0.09;
bsolution not prepared due to low solubility of the receptor.
Figure 6: Binding affinities of the studied receptors towards different carboxylates in DMSO-d6/H2O (99.5%:0.5% m/m).
As expected, the increase of water content in DMSO reduced due to the increased freedom of movement of the alkyl chains
binding affinities. The reduction was not uniform for the differ- as opposed to the fixed macrocycles (please see the Supporting
ent anions. According to Table 1, it was related to the hydrophi- Information File 1 for computational geometries), allowing for
licity of the ions – in the case of smaller and more hydrophilic more dynamic complexes to form. Such affinity was not
anions the reduction was more significant. A comparison be- entirely surprising as both, the open-chain receptor and the
tween all anions was possible with receptor MC009. The closed macrocycles, employed only aliphatic substituents in the
binding strength reduction with MC009 was 1.95 with acetate, 1 and 8 positions of the biscarbazolylurea moiety.
1
.75 with formate, 1.34 with lactate, and 1.22 with both pivalate
Preparation and characterization of sensor
prototypes
and benzoate.
Compound CZ016, being an open-chain receptor, competed The sensor prototype preparations were done similarly to our
well with the other macrocyclic structures. This was partially previously published method for an acetate-selective solid-con-
1908

Beilstein J. Org. Chem. 2020, 16, 1901–1914.
tact electrode [32]. Three macrocyclic receptors with different compare the influence of the ionophore upon the spectra, if any
ring sizes were chosen as ionophores: MC005, MC009, and could be observed. Examples of the recorded impedance spec-
MC012. Ion-selective membranes (ISMs) were prepared from tra are shown in Figure 7. The membrane bulk resistance and
poly(vinyl chloride) plasticized with bis(2-ethylhexyl)sebacate geometric capacitance were attributed to the high-frequency
(
DOS), which was chosen since it was considered to be suitable semicircle. The membrane bulk resistance and geometric capac-
for use in membranes with monovalent primary ions [5]. The itance were heavily influenced by the properties of the plasti-
compositions of the ISMs were 2 wt % ionophore, 50 mol % cizer, DOS (εr = 4.01 [34]), because it accounted for most of the
anion exchanger relative to the ionophore, 65 wt % DOS, and mass in all of the membranes. The bulk resistances were in the
3
2 wt % PVC. A control membrane was also prepared with no MΩ range, which was typical for PVC membranes plasticized
ionophore, 0.7 wt % anion exchanger, 66 wt % DOS, and with DOS (for the given membrane geometry) [33]. There was
3
1
3 wt % PVC. The dry weight of the membrane cocktails was some variation from sensor to sensor, which was attributed to
7% m/m. The sensors (Figure 2) were prepared with electrode variations in membrane thicknesses. The lack of any large semi-
bodies made from glassy carbon (GC) rods (diameter = 3 mm) circles or capacitive lines at lower frequencies showed that the
encased in PVC cylinders (outer diameter = 8 mm). The GC solid contacts had large redox capacitances [33]. The low-fre-
rods functioned as the electronic conductors and the PVC shells quency regions showed some differences between the different
were used to both control the amount of exposed GC surface membranes and MC005 stood out due to the presence of a
and to function as the substrate for the ISM to adhere to. partial semicircle that was absent in all other membranes. The
Poly(3,4-ethylenedioxythiophene) doped with chloride low-frequency region is associated with transport phenomena
(
PEDOT-Cl) was electropolymerized onto the exposed GC as a and this semicircle was attributed to slow ion transfer kinetics at
solid contact and provided improved potential stability in com- the ISM interfaces.
parison to a bare GC surface [33]. The membranes were applied
to the electrode bodies by drop-casting the membrane cocktails Potentiometric calibrations in sodium acetate solutions were
(
100 μL/electrode) and allowing the solvent to evaporate. Three used to determine the linear ranges (R2 ≥ 0.999), slopes, and
sensors were prepared with each of the membranes in an iden- detection limits of the sensors (Table 2). Examples
tical way and were conditioned in 0.1 M sodium acetate solu- of the calibration curves are shown in Figure 8. All of the
tions. The sensors were characterized using electrochemical measured slopes were sub-Nernstian (i.e., less negative than
impedance spectroscopy (EIS) and potentiometry. Please see −58.8 ± 0.4 mV/dec at temperatures of 23 ± 2 °C). The MC012
Supporting Information File 1 for detailed preparation and char- and MC009 sensors exhibited the best and second-best
acterization data of the sensor prototypes.
responses, respectively, according to all of the quantified char-
acteristics. The responses of the MC005 sensors were only
EIS measurements were used primarily to ensure that the solid slightly better than those of the control sensors. The poorer
contact had a sufficiently large redox capacitance and to response characteristics and greater standard deviations of the
Figure 7: Impedance spectra of sensors with each of the membranes. The spectra were recorded in 0.1 M sodium acetate in the frequency range
100 mHz to 100 kHz at open-circuit potential.
1909

Beilstein J. Org. Chem. 2020, 16, 1901–1914.
Table 2: The response characteristics of each membrane determined from the potentiometric calibrations in sodium acetate. The upper linear limit
was 10−1.11 M in all cases and activities greater than that were not measured. The uncertainties express standard deviations.
CTRL
MC005
MC009
MC012
slope (mV/dec)
log aLLLa
−52.14 ± 0.82
−2.77 ± 0.39
−3.74 ± 0.39
−54.59 ± 0.53
−3.24 ± 0.36
−4.16 ± 0.30
−55.77 ± 0.30
−4.00 ± 0.00
−4.94 ± 0.09
−56.56 ± 0.35
−4.00 ± 0.00
−5.12 ± 0.11
log aLODb
aLower limit of linear range; blimit of detection.
Figure 8: Calibration curves for each of the membranes. The calibrations were performed by diluting 0.1 M sodium acetate in half-decade steps until
the potentials of the sensors levelled out. The potentials were shifted along the y-axis to start at ΔE = 0.00 mV at log a = −1.11.
MC005 and control sensors could be attributed to their sensi- whereupon a clear decline in the potentials was present. A
tivity to chloride, which will be discussed in more detail later. steady decline could be seen in the potentials of the control and
MC005 sensors throughout the entire pH range. This again may
The pH sensitivity of the sensors was assessed by measuring be explained by their sensitivity to chloride, but the MC005
their potentiometric responses in acetic acid. The solutions were sensors exhibited a slightly stronger response than the controls
titrated from pH 3.3 to pH 10 with a mixture of sodium acetate did to the increase of acetate activity in the beginning of each
of equal concentration to the initial acetic acid solution and measurement. The deviations observed in the responses of the
concentrated sodium hydroxide. The activity of acetate was ex- MC005 and MC009 sensors below pH 6 could be attributed to
pected to increase between pH 2.8 and 6.8 due to the deproton- interference from chloride. Chloride was at least partially re-
ation of acetic acid (pKa = 4.76) [34], which should result in a sponsible for the deviations observed above pH 8 and pH 9, but
decrease in the measured potentials. The measured potentials interference from hydroxide and/or carbonate cannot be ruled
were then expected to remain constant after pH 6.8 since the ac- out.
tivity of acetate should no longer appreciably change with the
increase of pH. Structural differences between the different The selectivity of each membrane was evaluated with the sepa-
macrocycles were not expected to cause differences in pH rate solution method. The results for each interfering ion are
sensitivity. The results of these measurements are shown in shown in Table 3 and Figure 10. The carboxylates that were
Figure 9. The MC012 sensors were the least affected by the used in the determination of binding affinities in DMSO were
change in solution pH and their responses adhered well to the included to see how the affinities affect the selectivity of the
expected response with minimal interference. The MC009 ISMs towards acetate over other carboxylates. A few common
sensors adhered quite well to the expected response up to pH 9 non-carboxylate anions were also included in the selectivity
1910

Beilstein J. Org. Chem. 2020, 16, 1901–1914.
Figure 9: The influence of solution pH on the potential responses of the sensor prototypes (three sensors for each membrane composition). The mea-
surements were performed in 0.01 M acetic acid. Potentials have been adjusted to overlap the responses of sensors with the same membrane.
Table 3: Potentiometric selectivity coefficients of interfering anions (relative to acetate) determined using the separate solution method at ambient
temperatures. All anions were used as sodium salts. The uncertainties express standard deviations.
ion j
control
MC005
MC009
MC012
iodide
5.06 ± 0.07
3.60 ± 0.04
2.77 ± 0.09
1.48 ± 0.10
1.42 ± 0.07
0.50 ± 0.08
0.50 ± 0.11
−0.23 ± 0.31
1.91 ± 0.06
0.89 ± 0.06
2.44 ± 0.12
0.84 ± 0.11
1.61 ± 0.08
−0.16 ± 0.04
0.13 ± 0.08
−0.81 ± 0.28
1.28 ± 0.05
0.49 ± 0.02
2.13 ± 0.13
−0.52 ± 0.05
1.16 ± 0.05
−0.27 ± 0.00
0.12 ± 0.02
−1.84 ± 0.13
0.82 ± 0.02
0.04 ± 0.04
1.95 ± 0.07
−0.85 ± 0.02
1.12 ± 0.03
−0.30 ± 0.01
−0.26 ± 0.01
−2.22 ± 0.13
nitrate
benzoate
chloride
pivalate
lactate
formate
fluoride
measurements since the sensors should be able to discriminate The selectivity pattern for the control membrane followed the
interfering anions to be of practical use. Chloride was included Hofmeister series, as expected, with more lipophilic anions
primarily due to its presence in the filling solutions of the refer- causing more interference. The addition of any of the
ence electrode. As a result, it inevitably leaches into all of the ionophores improved selectivity in all but one case – the selec-
samples. The chosen non-carboxylate anions also represented tivity toward acetate with respect to pivalate was lower in the
different parts of the Hofmeister series. Chloride and nitrate are case of the MC005 membrane than it was for the control mem-
also common in real-world samples that such sensors might be brane. The interference caused by non-carboxylate anions de-
used to analyse. Fluoride and iodide represented halides at creased by up to four orders of magnitude by the addition of an
opposite ends of the Hofmeister series.
ionophore. Other carboxylates were discriminated to a lesser
1911

Beilstein J. Org. Chem. 2020, 16, 1901–1914.
extent. The selectivity towards chloride was important in under- abnormally large flow rate of chloride from the reference elec-
standing the responses observed during the calibration and trode during a single measurement. This affected the responses
pH-sensitivity measurements. Both the control and MC005 for sensors that were used.
sensors were more selective for chloride than for acetate. When
these sensors were combined with the reference electrode, The ISMs showed modest selectivities. However, incorporating
which had a 1 M potassium chloride outer filling solution, the any of the three receptors in the membrane as an ionophore
chloride interfered and caused drifting potentiometric responses clearly disrupted the Hofmeister selectivity pattern that could be
–
chloride gradually leached into the samples from the refer- seen in the control membrane where the lipophilicity of the
ence electrode over the course of the measurements.
interfering ion was the dominant factor in determining the
selectivity. All of the ionophores had a greater affinity for
Conversely, the lower sensitivity of the MC009 and MC012 pivalate than for benzoate based on the measurements per-
sensors to chloride explained their superior response character- formed in DMSO, but their corresponding ISMs were more
istics. The outer junction of the reference electrode was an selective for benzoate (log Poct/wat = 1.88) [34] than for pivalate
opening with an adjustable plastic plug, which can be used to (log Poct/wat = 1.48) [35]. However, the influence of the
empty the outer filling solution compartment, and the greater ionophores’ affinities for carboxylates were clearly visible from
standard deviations in the results could be explained by slight the suppressed selectivity towards anions without carboxylate
differences in the flow of chloride between measurements. groups (fluoride, chloride, nitrate, iodide), compared to the
These variations in flow rate were likely due to variations in the control membrane. Optimization of the control membrane's
position of the plug as a result of handling the reference elec- composition (e.g., reducing the amount of anion exchanger or
trode while setting up the measurements. This flow of chloride using another plasticizer) could improve its selectivity to some
resulted in narrower linear ranges, lower slopes, and poorer extent, but a disruption of the selectivity pattern based on the
detection limits due to the mixed responses to acetate and chlo- lipophilicity of ions in favor of a more hydrophilic primary ion
ride. The degree of deviation from the expected response in the would not be expected. However, ionophores can do just that as
pH sensitivity measurements decreased with the decreasing shown in Figure 10 by the reduced selectivity towards nitrate
sensitivity to chloride of the different ISMs. This also sup- and iodide [19].
ported the interpretation that chloride was at least partially re-
sponsible for the deviations. Finally, the greater standard devia- The sensors with the ionophore MC012 had the best perfor-
tions in the selectivity coefficients for fluoride were due to an mance in this study and those sensors also compared quite
Figure 10: Potentiometric selectivity coefficients of interfering anions (relative to acetate) determined using the separate solution method at ambient
temperatures. All anions were used as sodium salts. The uncertainties (error bars) express standard deviations.
1912

Beilstein J. Org. Chem. 2020, 16, 1901–1914.
favourably to some of the previously published ISEs for acetate (MSc) for designing the graphical abstract. Part of this work
by exhibiting a wider linear range, slope closer to theory, and a was carried out using the instrumentation at the Estonian Center
lower detection limit [36]. There was no overall champion in of Analytical Chemistry (http://www.akki.ee).
terms of selectivity when comparing the MC012 sensors to the
sensors from these earlier publications, as they outperform one Funding
another at discriminating different anions. Further optimization This work was supported by the Estonian Research Council
of the membranes could be done by trying different composi- grant PRG690, by the EU through the European Regional De-
tions (e.g., amount of ionophore, polymer:plasticizer ratio) and velopment Fund (TK141 “Advanced materials and high-tech-
plasticizers. While the sensors in this study were considered to nology devices for energy recuperation systems”) and by the
be acetate-selective for the purposes of characterization, the EMPIR programme (project 17FUN09, UnipHied) co-financed
sensors turned out to show the highest selectivity to benzoate by the Participating States and from the European Union’s
among the anions tested. These sensors could thus also be used Horizon 2020 research and innovation programme. JB grate-
as benzoate-selective sensors in applications where benzoate fully acknowledges funding from the Academy of Finland
would be the main analyte of interest.
(project no. 317829). VY gratefully acknowledges funding from
the Magnus Ehrnrooth Foundation.
Conclusion
To construct a functional chemical sensor, extensive interdisci- ORCID® iDs
plinary effort is required. A variety of problems must be
addressed in several development stages, all of which were
attended to in this work. The conceptual design of macrocyclic
anion receptors led to in silico complexation studies of biscar-
bazolylurea-type ionophores. Eleven novel methylene-bridged
macrocyclic receptor molecules were synthesized and character-
ized alongside a similar open-chain receptor and two additional
Alo Rüütel - https://orcid.org/0000-0003-4410-9932
Ville Yrjänä - https://orcid.org/0000-0001-5779-5201
Indrek Saar - https://orcid.org/0000-0002-9366-111X
Lauri Toom - https://orcid.org/0000-0001-9693-8180
Johan Bobacka - https://orcid.org/0000-0003-2346-9947
Ivo Leito - https://orcid.org/0000-0002-3000-4964
macrocycles. Experimental complexation studies with different References
carboxylates were carried out in two different solvent systems
to select three champions for constructing anion sensor proto-
types. Binding studies in solution suggested selectivity towards
acetate, however, this prediction did not hold true once the
ionophores were incorporated into sensors. Overall, the
constructed sensor prototypes showed modest selectivity
towards acetate, which increased as the ionophore's cavity size
increased. The sensors showed higher selectivity towards
the more lipophilic carboxylates pivalate and benzoate. In-
corporating any of the three chosen ionophores was shown to
reduce interference from more lipophilic non-carboxylate inter-
ferents (such as iodide or nitrate) when compared to a control
sensor.
1. Wolfbeis, O. S. Angew. Chem., Int. Ed. 2013, 52, 9864–9865.
doi:10.1002/anie.201305915
2
.
.
Albert, K. J.; Lewis, N. S.; Schauer, C. L.; Sotzing, G. A.; Stitzel, S. E.;
Vaid, T. P.; Walt, D. R. Chem. Rev. 2000, 100, 2595–2626.
doi:10.1021/cr980102w
3
Yang, L.-P.; Wang, X.; Yao, H.; Jiang, W. Acc. Chem. Res. 2020, 53,
198–208. doi:10.1021/acs.accounts.9b00415
4. Rehm, T. H.; Schmuck, C. Chem. Soc. Rev. 2010, 39, 3597.
doi:10.1039/b926223g
5
.
.
Mikhelson, K. N. Ion-Selective Electrodes; Springer Verlag: Berlin,
Heidelberg, Germany, 2013; Vol. 81.
6
Sanchez, G.; Espinosa, A.; Curiel, D.; Tarraga, A.; Molina, P.
J. Org. Chem. 2013, 78, 9725–9737. doi:10.1021/jo401430d
7. Hiscock, J. R.; Caltagirone, C.; Light, M. E.; Hursthouse, M. B.;
Gale, P. A. Org. Biomol. Chem. 2009, 7, 1781. doi:10.1039/b900178f
8
.
.
Martin, K.; Nõges, J.; Haav, K.; Kadam, S. A.; Pung, A.; Leito, I.
Eur. J. Org. Chem. 2017, 5231–5237. doi:10.1002/ejoc.201700931
Kadam, S. A.; Martin, K.; Haav, K.; Toom, L.; Mayeux, C.; Pung, A.;
Gale, P. A.; Hiscock, J. R.; Brooks, S. J.; Kirby, I. L.; Busschaert, N.;
Leito, I. Chem. – Eur. J. 2015, 21, 5145–5160.
9
Supporting Information
Supporting Information File 1
doi:10.1002/chem.201405858
Experimental part.
10.Ahmed, N.; Suresh, V.; Shirinfar, B.; Geronimo, I.; Bist, A.;
Hwang, I.-C.; Kim, K. S. Org. Biomol. Chem. 2012, 10, 2094.
doi:10.1039/c2ob06994f
[https://www.beilstein-journals.org/bjoc/content/
supplementary/1860-5397-16-157-S1.pdf]
1
1.Lee, S.; Hirsch, B. E.; Liu, Y.; Dobscha, J. R.; Burke, D. W.; Tait, S. L.;
Flood, A. H. Chem. – Eur. J. 2016, 22, 560–569.
doi:10.1002/chem.201503161
1
2.Zapata, F.; Caballero, A.; White, N. G.; Claridge, T. D. W.; Costa, P. J.;
Félix, V.; Beer, P. D. J. Am. Chem. Soc. 2012, 134, 11533–11541.
doi:10.1021/ja302213r
Acknowledgements
The authors thank the UT Proteomics Core Facility for help
with HRMS measurements. The authors thank Maris Taul
1913

Beilstein J. Org. Chem. 2020, 16, 1901–1914.
1
1
1
1
3.Gale, P. A.; Sessler, J. L.; Král, V.; Lynch, V. J. Am. Chem. Soc. 1996,
118, 5140–5141. doi:10.1021/ja960307r
License and Terms
4.Kim, S. K.; Sessler, J. L. Acc. Chem. Res. 2014, 47, 2525–2536.
doi:10.1021/ar500157a
This is an Open Access article under the terms of the
Creative Commons Attribution License
5.Zhou, H.; Zhao, Y.; Gao, G.; Li, S.; Lan, J.; You, J. J. Am. Chem. Soc.
2013, 135, 14908–14911. doi:10.1021/ja406638b
(http://creativecommons.org/licenses/by/4.0). Please note
6.Zahran, E. M.; Fatila, E. M.; Chen, C.-H.; Flood, A. H.; Bachas, L. G.
Anal. Chem. (Washington, DC, U. S.) 2018, 90, 1925–1933.
doi:10.1021/acs.analchem.7b04008
that the reuse, redistribution and reproduction in particular
requires that the authors and source are credited.
1
1
1
2
2
2
2
2
2
7.Schmuck, C.; Schwegmann, M. J. Am. Chem. Soc. 2005, 127,
3373–3379. doi:10.1021/ja0433469
The license is subject to the Beilstein Journal of Organic
Chemistry terms and conditions:
8.Busschaert, N.; Caltagirone, C.; Van Rossom, W.; Gale, P. A.
Chem. Rev. 2015, 115, 8038–8155. doi:10.1021/acs.chemrev.5b00099
9.Hein, R.; Beer, P. D.; Davis, J. J. Chem. Rev. 2020, 120, 1888–1935.
doi:10.1021/acs.chemrev.9b00624
(https://www.beilstein-journals.org/bjoc)
The definitive version of this article is the electronic one
which can be found at:
0.Chen, L.; Berry, S. N.; Wu, X.; Howe, E. N. W.; Gale, P. A. Chem 2020,
6
, 61–141. doi:10.1016/j.chempr.2019.12.002
1.Zhao, J.; Yang, D.; Yang, X.-J.; Wu, B. Coord. Chem. Rev. 2019, 378,
15–444. doi:10.1016/j.ccr.2018.01.002
2.Bakker, E.; Bühlmann, P.; Pretsch, E. Chem. Rev. 1997, 97,
083–3132. doi:10.1021/cr940394a
3.Bühlmann, P.; Pretsch, E.; Bakker, E. Chem. Rev. 1998, 98,
593–1688. doi:10.1021/cr970113+
4.Bobacka, J.; Ivaska, A.; Lewenstam, A. Chem. Rev. 2008, 108,
29–351. doi:10.1021/cr068100w
doi:10.3762/bjoc.16.157
4
3
1
3
5.Tshepelevitsh, S.; Kadam, S. A.; Darnell, A.; Bobacka, J.; Rüütel, A.;
Haljasorg, T.; Leito, I. Anal. Chim. Acta, in press.
doi:10.1016/j.aca.2020.07.024
2
2
2
2
3
3
6.Yawer, M. A.; Necas, M.; Sindelar, V. Tetrahedron 2016, 72,
2
943–2946. doi:10.1016/j.tet.2016.04.009
7.Dąbrowa, K.; Pawlak, M.; Duszewski, P.; Jurczak, J. Org. Lett. 2012,
4, 6298–6301. doi:10.1021/ol303065k
8.Dydio, P.; Lichosyt, D.; Zieliński, T.; Jurczak, J. Chem. – Eur. J. 2012,
8, 13686–13701. doi:10.1002/chem.201201909
9.Liu, Y.; Nishiura, M.; Wang, Y.; Hou, Z. J. Am. Chem. Soc. 2006, 128,
592–5593. doi:10.1021/ja058188f
0.Kadam, S. A.; Haav, K.; Toom, L.; Haljasorg, T.; Leito, I. J. Org. Chem.
014, 79, 2501–2513. doi:10.1021/jo4027963
1
1
5
2
1.Haav, K.; Kadam, S. A.; Toom, L.; Gale, P. A.; Busschaert, N.;
Wenzel, M.; Hiscock, J. R.; Kirby, I. L.; Haljasorg, T.; Lõkov, M.;
Leito, I. J. Org. Chem. 2013, 78, 7796–7808. doi:10.1021/jo400626p
2.Martin, K.; Kadam, S. A.; Mattinen, U.; Bobacka, J.; Leito, I.
Electroanalysis 2019, 31, 1061–1066. doi:10.1002/elan.201800790
3.Bobacka, J. Anal. Chem. (Washington, DC, U. S.) 1999, 71,
3
3
3
3
4932–4937. doi:10.1021/ac990497z
4.Haynes, W. M. CRC Handbook of Chemistry and Physics, 91st ed.;
CRC Press: Boca Raton, FL, U.S.A., 2010.
5.Chemicals Inspection and Testing Institute. Biodegradation and
Bioaccumulation Data of Existing Chemicals Based on the CSCL
Japan. Japan Chemical Industry Ecology-Toxicology & Information
Center Tokyo: Tokyo, Japan, 1992.
36.Umezawa, Y.; Bühlmann, P.; Umezawa, K.; Hamada, N.
Pure Appl. Chem. 2002, 74, 995–1099. doi:10.1351/pac200274060995
1914